feat: Add mlflow in model implementation mapping

Erouan50 · Erouan50 · commit 0dcd46a4f124 · 2024-08-28T09:03:24.000-04:00
The `mlflow` model format is supported by mlserver in kserver serverless
mode, and modelmesh too. Without the mapping, mlserver fails to load the
model.

Signed-off-by: Antoine Rouaze &lt;arouaze@mirakl.com&gt;
diff --git a/model-mesh-mlserver-adapter/server/server.go b/model-mesh-mlserver-adapter/server/server.go
@@ -369,6 +369,7 @@ func generateModelConfigJSON(modelID string, modelType string, uri string, schem
 		"sklearn":  "mlserver_sklearn.SKLearnModel",
 		"xgboost":  "mlserver_xgboost.XGBoostModel",
 		"mllib":    "mlserver-mllib.MLlibModel",
+		"mlflow":   "mlserver_mlflow.MLflowRuntime",
 	}
 
 	// set the implementation

Original file line number	Diff line number	Diff line change
`@@ -369,6 +369,7 @@ func generateModelConfigJSON(modelID string, modelType string, uri string, schem`
`369`	`369`	`"sklearn": "mlserver_sklearn.SKLearnModel",`
`370`	`370`	`"xgboost": "mlserver_xgboost.XGBoostModel",`
`371`	`371`	`"mllib": "mlserver-mllib.MLlibModel",`
	`372`	`+ "mlflow": "mlserver_mlflow.MLflowRuntime",`
`372`	`373`	`}`
`373`	`374`
`374`	`375`	`// set the implementation`